With recent advances in natural language processing techniques, a technique that makes information searching highly accurate is required. It is necessary to enable text searching at a level that matches the user's need and to feed back the searched result using a natural language processing technique.
An existing text base apparatus is configured to store as keywords words existing in a text base and searches the text by indexing these keywords.
However, in a search that uses keywords as indices, if a searched word has more than one meaning, the search is performed for a text that includes the semantics requested by the user, and the identification of the searched result depends on the user. Since a keyword setting depends on the text base creator, there is no consistency in the keywords and consistency in the searched result is not maintained.
Therefore, in existing text base apparatuses, searching a morpheme only (a minimum lingual unit having a meaning in a lingual system) has the defect that it cannot cope with multivocal words. Besides, since the text base is not linked with a dictionary, even if a new word is registered in a dictionary, the registration is not reflected in a prior text. Thus, consistency in the searched result cannot be maintained.